Performance & Benchmarks
Expected latency and scaling guidance for Dev SDK v1.0.4.
Expected performance
| Operation | P50 | P95 | P99 | Throughput |
|---|---|---|---|---|
| Tokenize | 5 ms | 20 ms | 50 ms | ≈ 2,000 req/sec |
| Detokenize | 3 ms | 15 ms | 40 ms | ≈ 3,000 req/sec |
| Tokenize (batch of 100) | 30 ms | 80 ms | 150 ms | ≈ 30,000 values/sec |
| Health check | < 1 ms | < 1 ms | < 1 ms | ≈ 10,000 req/sec |
Measurement conditions
Benchmarks measured against PostgreSQL on a 2-CPU / 4 GB pod with 1 million tokens already in the vault.
Scaling recommendations
| Scale | SDK pods | PostgreSQL | Storage |
|---|---|---|---|
| < 1M tokens | 1 pod | 1 pod, 2 CPU, 4 GB | 10 GB |
| 1–10M tokens | 2–3 pods | Dedicated PG, 4 CPU, 8 GB | 50 GB |
| > 10M tokens | Contact support | External managed PG | Sized to retention |
Above 10M tokens
If you anticipate more than 10 million tokens or sustained traffic above 200 req/sec, the Paid tier with replicated architecture and dedicated support is a better fit.
What Securelytix does NOT do
- We do not access your plaintext PII values
- We do not access your tokens or vault contents
- We do not share data with third parties
- We do not use your data for model training or analytics
- We do not see your API keys after issuance (only hashed validation digests)
- We track only aggregate request counts for licensing — never PII values or token contents